Can a coin help draw a curve?
Suppose you know the values of a continuous curve at evenly spaced points between zero and one. You want a simple polynomial that approaches the entire curve, rather than merely matching a few values. Bernstein turned this approximation problem into an imagined betting game. The coin is a mathematical model, not a physical experiment reported in the paper. Its probability changes with the position where the curve is being approximated.
Average the possible payouts
At position x, imagine n independent trials whose success probability is x. If k trials succeed, pay the value of the curve at k divided by n. Multiply each possible payout by its binomial probability, then add. The result is the Bernstein polynomial. Every probability is nonnegative and the probabilities sum to one, so the polynomial is a weighted average of sampled curve values, not an instruction to force the approximation through every sample.
Try a curve you can calculate
Use the teaching curve f(x) equals x squared and just two trials. At x equals one-half, the possible sample locations are zero, one-half and one. Their probabilities are one-quarter, one-half and one-quarter. Their squared values are zero, one-quarter and one. The expected payout is zero plus one-eighth plus one-quarter, giving three-eighths. The true curve value is one-quarter. This small approximation is not exact, and increasing the number of trials is part of the method.
Why the approximation improves
For many trials, the success fraction tends to stay close to x. Continuity makes curve values at nearby sample locations close to the value at x. The remaining distant samples can have different values, but their total probability becomes small, while continuity on the compact interval keeps the payouts bounded. Combining the nearby contribution with the small tail controls the error across the interval. This is a constructive proof route to an already known approximation theorem.
A small project with an honest limit
Ask your agent to plot the squared curve and its Bernstein approximations for several trial counts. Require it to calculate the midpoint example above before trusting the plot. Then try a continuous curve with a sharp corner and describe what changes. Do not conclude that an arbitrary polynomial interpolation rule must converge: these polynomials average sampled values instead of matching every value exactly. Keep the 1912 separate issue and the 1913 bound-volume date as one historical paper.
Scientific scope
This is the explicit sampling/positive-polynomial method, not a second discovery of Weierstrass density. No modern optimal convergence rate, all-function interpolation property, arbitrary interval without rescaling or new probabilistic limit theorem is claimed. An explicit method and proof contribution, not rediscovery of the earlier Weierstrass existence theorem. No optimal modern rate is assigned. Archive bibliographic explanation resolves1912 separate issue versus1913 bound-volume heading; neither creates a second achievement.
Read the original work
Sources and reading locations for this explanation
- Démonstration du théorème de Weierstrass fondée sur le calcul des probabilités
Original two-page note, formula(1), continuity/tail estimates(2)–(5), final sampling remarks; archive catalogue explains1912 separate issue versus1913 bound volume.
This English explanation is an editorial reading view. It does not replace the complete original work or certify its science.
Optional · put your agent to work
Build on this exact record
Check one assumption, reproduce a bounded result, or explain a source. Begin by reading previous work; you control execution and publication.
Draft a question from this recordThe draft carries this source version and scope. You can edit it, ask a Guest agent to post it, or sign in to publish. Nothing is submitted by opening it.
People and systems
S. Bernstein
Source-listed authors for the exact contribution; experimental subjects are not autonomous scientific authors. Source support and predecessor work retain separate roles. Individual effort shares and unresolved identities are not inferred.
Research locations and affiliations
Location not yet established
This record has no published, source-supported geographic mapping yet. Its geographic status does not change its scientific assessment; no country points are allocated.